It’s way, way too late for me to be starting a Hive post right now. 3:30a.m., and @guthrie’s half-brother is sleeping over tonight. They’ll both surely be up early and neither are known for their discreet consideration in noise and volume. I’m sure I’ll pay for this decision, but it’s one of those nights that I have something to share. Over the years I’ve done reoccurring posts on my black and white photography, and going back even further, I hosted a content called Trial by Comics. I think I have to make some of my AI generated art a regular part of my personal blog because I’m not getting used to it at all.
There’s some debate about where this technology is going. Some think the fact that it’s getting so good will eventually lead to us being less impressed, or over-exposed to, “perfect”, art. I can see that, honestly, but just when I think I have the trajectory of this revolution figured out, it changes. MidJourney specifically has a very distinct look to its prompt results. They were epic, painterly and well composed, but they were still identifiable as MidJourney, without question. With the latest version 4 release, that’s no longer true.
Illustration and styles have been improved making it possible to generate art that spans a wide range of influences. Concept art, 2D illustration, even architectural drafts are possible with only a few integral prompts. All day I’ve been getting lost in these results. I save many of my best generated pieces of art but lately, almost everything is getting saved. The latest version has drastically improved faces, symmetry, light, anatomy and style. I’ll be releasing my interview with @gric soon where we discuss much of this.
What I’m most astounded by is the exponential evolution of the algorithm. It was about six months ago that I discovered MidJourney and it was almost impossible to get a figure or form generated with accurate appendages. Legs and arms were like spaghetti with no discernible beginning or end. Now, the only area I’m still seeing conflicts are in eyes [occassionaly] and fingers/hands. I’m predicting these will both be addressed in version 5, and I really do think at that point, it’ll be game over unless human-generated art is specifically what clients and companies want.
I doubt it'll be game over dude.
Wow, those are some impressive pieces. I need to get back into playing around with it. Putting words together has been considered an art for millennia, so I don't think it is too far off the mark. Just a different form. We've seen shifts like this over and over again. That's my opinion anyway.
Oh wow that’s amazing picture